i got dynamanic ip address at the moment. and i wondering to change to static ip so my ip address never changes. do you think my problem will solve then?

For what it's worth, I had an identical experience with my ALIX 3D2.

By default, the BIOS baud rate is at 38400.  I set minicom to 38400, and fired it up.  When the machine POST-ed, I typed "S" to enter the BIOS setup.  I set the BIOS baud rate to 9600 so that it would match the BSD default (at least that's my understanding of the default).  I saved the settings.  Then I reset the minicom baud rate and power cycled the ALIX.  The BIOS test was readable but I got exactly the same double characters until the kernel loaded and started booting.  Then the double character anomaly was gone.  So the double character business seems isolated to the boot loader.  Does anyone know why this might be happening?  Thanks in advance

@bob76535:

I haven't messed with squid yet. Can it be set to proxy some sites and not proxy others? I don't want it to cache the sites we work on as the users are constantly changing them and need to see the changes immediately. I would need to set it to skip our entire IP range (254 addresses). As long as that wasn't a problem, that sound like a good plan.

I had read elsewhere that the processor speed was not really important with pfsense but bus speed and memory speed was important. Is that not true?

Thanks

Bob

Yes, and even with squid they could see changes immediatly.

Dont know about the memory bus speed, many of us are running with Pentium 3 -level stuff. As am I for example, an old isp1100 with two intel network cards and P3 750MHz. 1Gb of ram and its running good and fast, even with squid on it.

I was able to fix it by one of two things.
Forcing the NIC to 100mbps seemed to work okay. It would flake sometimes. And using a 100mbps NIC also works, which seems to work the best out of the two solutions.

I don't think its an issue with pfsense, its the modem that doesn't play nice with the card. Thank you for your help.

I can't see any reason why it wouldn't work. However I have no experience with that board so can't say it will work. (I  suspect its pretty new and new stuff sometimes needs some driver tweaks before it will work.)

If you are looking for a mini-ITX dual LAN board that someone has found to work with pfSense search the forums for some positive reports on a MSI board. I think its model number begins with IM.

I think "disconnect" effect is solved for me. I had it when I connect some clients directly to lan of pfsense or switch was problematic (after replace no problems). I am now playing width 1.2.2 but whenever I play width traffic shaper wizard I have xfiles in some points. It wasn't like this in 1.2.0 at least for my systems.

Thanks for your reaction, Cry Havok. I've searched, found and read your earlier comment, but it doesn't speak about the 8139D. I've sold some systems that are suited for using pfsense, untill now only with Intel NIC's. Realtek isn't perfect for me either. But the most basic idea of network adapters is: do they work without errors? The Realtek 8139D NIC's I'm testing are working without errors. The price difference between Intel one 1000GT and a one Realtek NIC is approximately $24/€19 (In the Netherlands). For a lot of people that makes the difference between a pfsense-based firewall and, lets say, a cheap Linksys router with embedded firewall. But I agree that for maximum performance the only solution is an adapter with Intel chipset.

Had similar issues myself, I recently replaced my old T22 Thinkpad laptop that was running PFSense 1.2, with the new 2.0 Alpha release on a faster T23 laptop.  They have onboard Intel NICs, and on the T23 I tried using a 3COM 656 cardbus…it was working, but sporadic disconnect issues.  I yanked it, slapped in a DLink PCMCIA NIC...reset PFSense to use it..and all is good.

Just the one install (home FW)
I got everything put together.  Found a spare 256MB card and loaded it on that.
Restored my current pfsense config onto it and it took it fine (it was one rev before the embedded version that I just installed)

I'm up and running!  Love it!

@g4m3c4ck:

@Seth:

Can anybody say AirPort Express.

I would love a device that I could load pfSense on that would be a direct replacement for an AirPort Express.

Good point! I found this little guy right  here http://www.slashgear.com/toradex-robin-atom-cpu-on-credit-card-sized-mainboard-2735953 that make me ponder what else I could use it for  ;D

The cost of the device is double what's listed on slashgear as you need the carrier board to actually make it do anything.  At that price, a cheap much faster PC is a better option.

–Bill

I've got a couple jetway systems.  I love them.  infact, they make a mini-itx that takes a 4 nic daughter board that's pretty cool too.  I've never had a problem with jetway products but i have a friend who received a DOA board…he just returned it and hasn't had a problem since.  That happens with ALL manufacturers though.

hm, it seems i have mixed up my machines and posted dmesg from smp kernel machine. but as i was about to check it on the second machine with uniprocessor kernel, it was not able to recognize the onboard realtek nic just like when runinng with smp kernel (same re0: PHY write failed errors)

i tried a little bit of googling and found out it is known problem with freebsd 7.0 and 7.1. so i dont know this is the proper place to ask for support regarding bsd kernels, but maybe some kernel dev is around here and willing to help.

ECC is more for hardware reliability - a standard PC with non ECC memory might blue screen if memory is corrupted but a server with ECC memory will detect a single bit error and correct it. In the old days a server with ECC memory would still blue screen (windows) on a double bit fault where as now-adays the DIMM will more likely be marked off-line.

Alternatively buy two R200s and set up CARP.

I agree with Cry Havok, though my personal experience is that well-engineered PSUs are much less likely to fail than hard drives, it's still a single point of failure with moving parts. And mirroring is probably cheaper than SSD in a real enterprise-ready context anyway; I'm not even sure SSD has dropped yet in enterprise hardware, I think it'll be another generation or two before we see it become widespread there. Anyway, if you really want to pursue the cheap Dell with SSD option, it is possible. Get the 2x PCIe 8x riser, drop in an Intel dual or quad adapter and in the free slot plug in a PCIe -> mini PCIe adapter and a mini PCIe based SSD.

Intel's quad adapter is a PCIe 4x device. That's 1GB/s bandwidth, which is just fast enough for all 4 adapters running full steam in both directions at once. Okay, you might be slightly limited, but in the real world, you're never going to have 8Gb/s running through them.

The NC364T is based on the Intel 82571, from the HP spec sheet. This chip is supported by FreeBSD's em driver, so I see no reason for it not to work.

The NC7782 isn't explicitly listed, but it contains a Broadcom 5704, which is listed on the bge manpage as a supported chip and it should work (though the Intel chip is still better). Have you tested it and found it not to work?

Those Broadcom NICs suck (everywhere, not just under pfSense).  I just had a conversation about that yesterday afternoon with a couple Dell reps during a presentation on their upcoming servers.  I mentioned how they just start to drop packets under load and then eventually take out the system.  I use the 2900/2950 systems for my ESXi boxes and I've disabled all the Broadcom parts and replaced them with Intel Server NICs.

With modern wear-leveling algorithms you "should" not hit the write-limitation.

Assume a disk of 4 GB.
If you write constantly with 50 MB/s to the disk you would write every ~80 seconds to the same cell.
4096/50 = 81.92
Assume conservatively that you can write 100'000 times to a single cell on an expensive card (some claim to be able to write 1'000'000 times)
100'000 * 80s = 8'000'000s
8'000'000s / 60 = 133'333.3 min
133'333.3 min / 60 = 2222'.2 hours
2222.2 hours /24 = 92.6 day

So if you write constantly with the maximum speed to a disk with a conservative assumption of maximum 100'000 writecycles the card would start to show first signs of failure after 93 days.

Of course if you take a 16 GB card (*4),
"only" 10 MB/s (*5, still a lot of disk traffic…)
and 1'000'000 writecycles (10)
you have 93 days 1054 =18518 day
18518 / 365 = 50.7 years.

But the more expensive cards are about the same price as a harddisk ;)

What version of pfSense? If you haven't tried 1.2.3 please do so before you report back. You can download a snapshot from http://snapshots.pfsense.org/FreeBSD7/RELENG_1_2/?C=M;O=D

If you have the problem with pfsense 1.2.3 please post the startup output and a list of the PCI devices in the system. I presume you have at least one working interface so you can attach via the web GUI. If so, you can get the requested output by using the Diagnostics -> Command menu item and give the command dmesg ; pciconf -l -v.